Recommend "icc -fp-model precise" if doing "make check".

[jkbonfield]

Without this we get 24 test failures, as the default -fp-model is a bit fast and loose. (Ideally our tests wouldn't object so much about minor fluctuations in the least significant digits, as that's not really the thing we're trying to validate.)

An example, from the first test failure (icc 2021):

                        < 17    2041    .       G       A,<*>   0       .       
DP=7;I16=0,0,4,3,0,0,276,11066,0,0,420,25200,0,0,137,2895;QS=0,1,0;VDB=**0.80233**;S
GB=-0.636426;MQSBZ=0;FS=0;MQ0F=0        PL      223,21,0,223,21,223
                        ---
                        > 17    2041    .       G       A,<*>   0       .       
DP=7;I16=0,0,4,3,0,0,276,11066,0,0,420,25200,0,0,137,2895;QS=0,1,0;VDB=**0.802329**;
SGB=-0.636426;MQSBZ=0;FS=0;MQ0F=0       PL      223,21,0,223,21,223

[pd3]

I had a code in test.pl that was making it more robust to such floating point differences. Can't find it though, not sure where it ended...